Flow
How It WorksFlow by ELDFlow by TextFlow by WalletFlow InsideIntegrations
Solutions
Enterprise FleetsGrowing FleetsOwner Operators
Resources
Live Diesel PricesFree Fuel PlanAPI Docs
PricingBook a demo →Sign in →
Buyer guide

Detecting fuel card fraud and fuel theft.

Fuel card fraud is caught at the transaction level, by comparing every purchase against what the truck was actually doing at that moment. Card controls alone will not find it, because a compromised card used within its limits at an approved merchant looks exactly like a legitimate fill. What finds it is running each transaction against the trip, the tank, the clock and the driver's history: gallons exceeding tank capacity, two fills too far apart for the time between them, the same card twice at one station minutes apart, purchases outside the driver's duty window, fills with no active load assigned, and purchases well off the planned route. ValveRide Flow runs eight such detectors on every transaction a fleet posts. The output is a review queue, not an accusation. Most hits have innocent explanations, which is the point: you want the short list of things worth a look, not a verdict.

Updated August 2026 · 8-minute read

The eight detectors, and what each one looks for

Every one of these runs on every transaction a fleet posts. Each surfaces transactions to review rather than reaching a conclusion.

Overfill

  • Gallons purchased exceed what the truck's tank can physically hold
  • Classic signal for fuel going into something other than the tank
  • Innocent explanations: wrong tank capacity on file, a reefer or auxiliary tank filled on the same transaction

Impossible distance

  • Two purchases too far apart for the time elapsed between them
  • The truck could not have physically driven between the two stops
  • Innocent explanations: a team operation, or a card legitimately shared during a swap

Double swipe

  • The same card at the same station within minutes
  • Often a second vehicle filled on one driver's card
  • Innocent explanations: a genuine retry after a declined authorization

After-hours fueling

  • A purchase outside the driver's hours-of-service duty window
  • Cross-referenced against ELD status rather than a fixed schedule
  • Innocent explanations: a logging error, or a legitimate stop during off-duty time on a long layover

Off-dispatch fueling

  • A fill with no active load assigned to that truck
  • Meaningful because personal use tends to happen between loads
  • Innocent explanations: deadhead moves and yard repositioning not captured as dispatched trips

Off-route fueling

  • A transaction well off the planned corridor for the active load
  • Distinct from a small detour: the threshold is set to ignore ordinary variation
  • Innocent explanations: a reroute, a parking or scale closure, or a driver break requirement

Repeat-station drift

  • Habitual stops that consistently bypass the optimized plan
  • Usually a compliance and coaching signal rather than a fraud one
  • Included because the two look identical on a spend report and different in transaction data

Price drift

  • Price paid diverges from your contracted rate or the local rack
  • Catches discounts that silently stopped applying, which is more common than theft
  • Often the detector that pays for itself first, because it finds a broken contract feed

Why a fuel card report will not find this

Controls are not detection

  • Per-driver limits and merchant restrictions stop obvious misuse and nothing subtle
  • A compromised card used within its limits at an approved truck stop is indistinguishable from a real fill
  • Controls are necessary and not sufficient, which is why fraud persists at fleets that have them

Context is what makes it visible

  • Detection needs the trip, the tank, the hours-of-service clock and the driver's history alongside the transaction
  • That data lives across the ELD, the TMS and the card feed, not in any one of them
  • Bringing them together is the actual work, and it is why fuel optimization and fraud detection tend to ship together

A queue beats an alarm

  • High-confidence-only alerting misses the patterns that matter, which are cumulative
  • Firing on everything trains people to ignore it
  • The useful output is a ranked short list a human reviews weekly, with the reason for each flag shown

How this works in ValveRide Flow

Anomaly Watch

  • All eight detectors run on every transaction your fleet posts, on Growth and Enterprise
  • Each flag shows why it fired and the transaction behind it, so a manager can resolve it in a minute rather than investigate from scratch
  • Built on data Flow already has for optimization: the planned route, the tank state, the hours-of-service clock and your contract rates
  • Presented as things to look at, never as accusations, because most flags have innocent explanations and treating drivers as suspects is how these programs die
  • Price drift in particular tends to surface broken discount feeds, which is a billing problem worth catching whether or not anyone is stealing
Questions, answered

Common questions.

What software can detect fuel card fraud or fuel theft in a trucking fleet?

Fuel card fraud is caught at the transaction level, by comparing every purchase against what the truck was actually doing at that moment. Card controls alone will not find it, because a compromised card used within its limits at an approved merchant looks exactly like a legitimate fill. What finds it is running each transaction against the trip, the tank, the clock and the driver's history: gallons exceeding tank capacity, two fills too far apart for the time between them, the same card twice at one station minutes apart, purchases outside the driver's duty window, fills with no active load assigned, and purchases well off the planned route. ValveRide Flow runs eight such detectors on every transaction a fleet posts. The output is a review queue, not an accusation. Most hits have innocent explanations, which is the point: you want the short list of things worth a look, not a verdict.

How common is fuel theft in trucking really?

Common enough that most fleets above a handful of trucks find something when they first look, but the something is usually not dramatic. Compromised card numbers, friends-and-family fill-ups, and personal-vehicle fills are the recurring patterns. Just as often the first real finding is not theft at all, it is a discount that stopped applying months ago, or a tank capacity recorded wrong on a truck. Both cost money and neither shows up on a spend report.

Will this create a lot of false positives?

Some, deliberately. A detector tuned to fire only when it is certain will miss the cumulative patterns that matter most, and those patterns are the ones that add up. The design goal is a short ranked review queue rather than either an alarm or a verdict, with the reason for each flag visible so a manager can dismiss an obvious explanation quickly. Thresholds like off-route also require sustained signal rather than a single ping, which removes most of the noise.

Do we need to accuse a driver to act on a flag?

No, and you generally should not start there. Most flags resolve into a data problem or a routine explanation. The productive posture is to treat the queue as a list of things worth understanding: a wrong tank capacity gets corrected, a broken discount feed gets fixed with the vendor, a repeat-station pattern becomes a coaching conversation, and only a small residue looks like anything else. Fleets that open with accusations tend to stop using the tool.

What data do you need from us to run detection?

The fuel card transactions and enough operational context to judge them: the trips, the truck and tank details, and hours-of-service or position data where available. In practice this is the same data Flow already uses to build fuel plans, which is why detection comes with optimization rather than as a separate product. A fleet running Flow for planning has the inputs by definition.

Is this available without buying fuel optimization?

Anomaly Watch is part of Growth and Enterprise rather than a standalone product, because it depends on the same route, tank and contract data the optimizer builds. If detection is your only interest, that is worth saying in a demo so the conversation stays on it.

Related: getting discounts onto every load, how to audit a savings claim, how Flow works.

Run the detectors on your own transactions.

A 30-minute demo can walk your recent fuel card data through Anomaly Watch and show what the queue looks like for your fleet, including the boring findings that turn out to be worth money.